[Kst] Tag and branch

[George Staikos]

On Monday 22 November 2004 20:09, Barth Netterfield wrote:
> > Tomorrow I will tag 1.0 and create packages, probably somewhere around 3
> > or 4 pm.  We can do any testing in the early evening and I can upload it
> > to the FTP server after it looks OK.
>
> I just did cvs2dist here and we are installing on a couple BLAST machines
> to test.  Both realtime work, and some data analysis work from ascii.

  Excellent.  Matt, do you have any spec file updates?

> > I also need to discuss with the i18n team on how to arrange i18n issues
> > with two branches since they're only "capable" of dealing with
> > KDE-specific tags and branches.
>
> What does this mean?

  I'm not sure yet.  I need to find out. :-)

> > Finally, when we unfreeze, I would like to propose that we actually not
> > do new development for a week or two, just bug fixing, and more
> > importantly, developing a testing framework and implementation.  Included
> > in this would be some reworking of compilation structure of Kst to
> > facilitate this, and facilitate the building of a Kst library for other
> > C++ apps.  The equation testing tool is phenomenally powerful, and I
> > think it shows that we: a) really need this in other areas (PSDs,
> > plotting code, histograms, plugins, UI, labels, and more)
> > b) have many undiscovered bugs that don't hit our own common usage cases
> > c) have no way to systematically detect regressions
> > We need a test suite, and I don't see us doing it unless we force
> > ourselves to work on it.  It's more important to stabilizing Kst than
> > bugfixing, in my opinion.
>
> Given that we are going into doing math in production environments, I
> reluctantly agree.  I don't have any experience in writing such things,
> however.  I think we need a plan of attack, with some tasks we can all work
> on.

   I can start putting this together and getting a framework that's easy to 
dig into.  I have part of it here on my laptop but it's not safe to commit 
yet.  I was hoping to have something automated that would report at the end 
of each week to the list with the results of the test run.  I can host a 
machine that automatically pulls down Kst, rebuilds it, and runs the 
regression suite with profiling enabled so that we even get % code coverage 
statistics.  As code coverage drops, we will know that we need to add more 
testcases if we haven't already.

   I will commit a template for creating testsuites and eventually I will 
refactor the Kst build format to make it very easy to add new testcases in 
other areas.  What I'm using right now is quite a mess...

-- 
George Staikos
KDE Developer				http://www.kde.org/
Staikos Computing Services Inc.		http://www.staikos.net/